Kernel Based Image Classification

Authors:
Olivier Teytaud;David Sarrut
Affiliations:
-;-
Venue:
ICANN '01 Proceedings of the International Conference on Artificial Neural Networks
Year:
2001

Citing 11
Cited 1

Computer graphics: principles and practice (2nd ed.)

Computer graphics: principles and practice (2nd ed.)
The nature of statistical learning theory

The nature of statistical learning theory
Matrix computations (3rd ed.)

Matrix computations (3rd ed.)
Playing billiards in version space

Neural Computation
Neural Networks for Pattern Recognition

Neural Networks for Pattern Recognition
A New Multi-Class SVM Based on a Uniform Convergence Result

IJCNN '00 Proceedings of the IEEE-INNS-ENNS International Joint Conference on Neural Networks (IJCNN'00)-Volume 4 - Volume 4
Using machine learning to improve information access

Using machine learning to improve information access
Solving multiclass learning problems via error-correcting output codes

Journal of Artificial Intelligence Research
Comparing support vector machines with Gaussian kernels to radialbasis function classifiers

IEEE Transactions on Signal Processing
The sample complexity of pattern classification with neural networks: the size of the weights is more important than the size of the network

IEEE Transactions on Information Theory
Support vector machines for histogram-based image classification

IEEE Transactions on Neural Networks

A framework for incorporating class priors into discriminative classification

PAKDD'05 Proceedings of the 9th Pacific-Asia conference on Advances in Knowledge Discovery and Data Mining

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this study, we consider low-level image classification, with several machine learning algorithms adapted to high dimension problems: kernel-based algorithms. The first is Support Vector Machines (SVM), the second is Bayes Point Machines (BPM).We compare these algorithms based on strong mathematical results and nice geometrical arguments in a feature space to the simplest algorithm we could imagine working on the same representation. We use different low-level data, experimenting lowlevel preprocessing, including spatial information. Our results suggest that the kernel representation is more important than the algorithms used (at least for this task). It is a positive result because it exists much more simpler and faster algorithms than SVM. Our additive low-level preprocessings only improved success rate by few percents.